CLGE SC2012_Alblas. the NETHERLANDS__Archaeological Visibility Analysis With GIS

7/28/2019 CLGE SC2012_Alblas. the NETHERLANDS__Archaeological Visibility Analysis With GIS

1/7

1

1. INTRODUCTIONThe popularity of the usage of Geographical

Information Systems (GIS) to analyse and explore past

landscapes has grown rapidly over the last few

decades. The purpose of visibility analysis in

landscape archaeology is to explore the visual

organisation of features across a landscape (Kay and

Sly, 2001). Visibility is of major importance to how

humans relate to and interpret the landscape. Peopleoften describe a place based on the visibility (Wheatley

and Gillings, 2001). Visibility analysis is therefore an

important element in the interpretation of the

landscape for understanding past societies. Visibility

analyses can help analyse the spatial distribution of

features in the landscape or help answer the question

why a particular site was in a particular place

(Wheatley and Gillings, 2001).

2. THEORETICAL BACKGROUND2.1 Visibility analysis

Different GIS packages contain different visibility

analysis tools. This research focusses on ArcGIS

version 9.3 (with Spatial Analyst and 3D Analyst

extensions) which is used at the Department of

Archaeology at the moment. Focus will be on theViewshed Tool which generates an overview of the

visibility of the area from one or more viewpoints.

The most common viewshed analysis is the binary

viewshed analysis. The output is a raster that shows

the cells that are visible and cells that are not visible

from one or more viewpoints. The raster is created by

producing a line-of-sight analysis from the viewpoint

to each cell in the height input model; a Digital

Elevation Model (DEM). Cells that are not blocked by

elevation heights of other cells are classified as visible

and are assigned the value 1. All other cells areinvisible and are assigned the value 0. Each raster

contains cells with either the value 1 or 0 and is

ABSTRACT:

Spatial information is of great importance to archaeologists. Many archaeologists have adopted Geographical

Information Systems (GIS) in their studies because GISs can make managing, interpreting and visualising

geographical information easier. Visibility issues are of great importance to archaeologists, especially in landscape

studies. A GIS can be used to analyse visibility, but the available tools have been criticised by archaeologists and GIS

researchers. The standard analysis that can be conducted using GIS does not always represent reality well enough and

does not take all factors into account that could have an impact on visibility. In this study the factors influencing

visibility have been researched. Vegetation is one of the factors mentioned in literature that can influence visibility

significantly. Several approaches have been proposed in literature, ranging from very simple methods to methods

taking the possibility of seeing through vegetation into account. Uncertainty is also an issue in visibility analysis. The

output of the visibility analysis is often seen as the truth whereas in reality the analysis is the result of modelling and

is therefore one of the possible outcomes. Several approaches are named in literature. This study focusses on

archaeological visibility studies in general, but a study area in Scotland has been used as a study case. The best

approaches for the study area have been chosen to incorporate vegetation and uncertainty by developing tools that

can be used in a GIS. The resulting tools have been used to analyse the visibility with the incorporation of factors

influencing visibility. The analyses have been useful for the archaeologists to assess the visibility of and/or frompoints of interest but also to see the influence of factors on visibility. The tools give a better insight in the factors

influencing visibility and invite archaeologists to think through their questions about the significance and meaning of

articular sites.

ARCHAEOLOGICAL VISIBILITY ANALYSIS WITH GIS

Paper based on a thesis carried out for the Department of Archaeology, University of Glasgow, and submitted as a

requirement for the degree Bachelor of Built Environment (Geodesy/Geoinformatics), HU University of Applied

Sciences Utrecht the Netherlands.

Linda Alblas

KEYWORDS: GIS, landscape archaeology, uncertainty, vegetation, viewshed, visibility analysis


2/7

2

therefore called binary. Variations to the binary

viewsheds exist, such as cumulative viewsheds where

viewsheds from different viewpoints are added.

2.2 Limitations

Visibility analysis is complex. Wheatley and Gillings

(2000) classify the issues relating to visibility analysis

into three categories:

Pragmatic - pragmatic issues are those whichapply to both GIS and non-GIS based visibility

studies

e.g. vegetation, human perception and temporal

changes

Procedural procedural issues refer toconcerns that arise as a product of using GIS

for visibility analysis

e.g. DEM accuracy and the undifferentiated natureof the viewshed (binary output)

Theoretical theoretical issues are those whicharise from debates in the humanities (e.g.

geography)

The list of issues relating to visibility analysis is long.

Several methods, algorithms and tools have been

proposed by several authors to overcome these

limitations and to enrich existing methods. Focus of

this research will be on exploring the possibilities ofincorporating vegetation and visualising uncertainty

of the viewshed outcome caused by DEM uncertainty

and human perception for a study area in Scotland.

2.3 Research question

This research focusses on addressing and evaluating

the limitations of not incorporating vegetation and the

inability to model uncertainty caused by human

perception and error in the DEM. It also focusses on

applying enriched forms of viewshed analysismentioned in literature to overcome these limitations.

The main research question is:

Can the incorporation of vegetation and visualisation

of uncertainty lead to better visibility analysis for

landscape archaeology and for the SERF project in

particular?

3. STUDY AREAThe SERF project serves as a study area for this

research. SERF stands for Strathearn Environs & Royal

Forteviot. Strathearn is a district in Scotland (part of

the council area Perth and Kinross). Forteviot is a

village in Strathearn. Archaeological research has been

conducted in Strathearn and in and around Forteviot

in particular. Archaeologists from the University ofGlasgow are involved in this long-term project that has

been running since 2006. A 10m DEM, which

represents the bare terrain, is available for the study

area. The DTM has been created by Ordnance Survey

based on their contour data. Two study periods where

visibility is of importance have been used as study

cases; the Neolithic period, researching the visibility

from Neolithic ceremonial structures and the Iron Age,

researching the visibility from hill forts.

Figure 1. Location of the SERF project (source map: ArcGISExplorer topography basemap)

Neolithic

Cropmarks have revealed enclosures and ceremonial

structures near the village Forteviot in Scotland. It is

likely that the surrounding area of Forteviot was

largely wooded in the Neolithic. Questions raised

relating to visibility are: were the two Neolithic

complexes intervisible? What was the visibility from a

Neolithic complex?

Iron Age

The Iron Age in Scotland began around 700 BC and

ended around 500 AC. Forts were built on hills and

would have been built predominantly as centres of

power in the Iron Age. Most of these forts stand on

the northern slopes of the Ochils, with views across

Strathearn. Questions raised relating to visibility are:

what was the visibility from a fort? Which other hill

forts were visible from a particular hill fort?


3/7

3

4. CONDITIONS FOR THE TOOLSMost archaeologists use GISs for basic analysis of

gathered data. More sophisticated analysis tools, such

as visibility tools, are being used in archaeology but

nothing has been done on viewshed analysis for the

SERF project. For most archaeologists, a GIS is one of

the tools to explore past landscapes and study past

human societies. Not all archaeologists use a GIS everyday. A few conditions have been formulated in order

to develop tools that are useful to archaeologists.

The main conditions are that the tools need to be user

friendly and relatively easy to understand. Therefore

straightforward tools with a tutorial need to be

provided. Another condition is that the tools need to

be integrated with ArcGIS 9.3 because this is used at

the department.

5. METHODOLOGY: improving the DEMinput

The DEM available only represents the terrain. Man-

made structures and vegetation also have an influence

on visibility however. The study area was not urban

and only a few prominent man-made structures were

present that will serve as viewpoints in the viewshed

analysis. Man-made structures are therefore not

considered to be incorporated. Vegetation, however,

was present and archaeologists are interested in

exploring the effects vegetation could have had on

visibility.The inability of incorporating vegetation in visibility

analysis has been a limitation to model visibility

accurately. In flat areas the correct modelling of

vegetation is of high importance when modelling an

area for visibility purposes, especially in wooded

areas. In steeper areas both vegetation and topography

are important (Ashton, 2010). Including vegetation in

the DEM would contribute to obtain a visibility

analysis as close as possible to reality. Vegetation can

be considered as an accident of the terrain model. A

vegetation elevation model can be created that can be

added to the DEM. This method is mentioned in

ArcGIS 9.3 Desktop Help and it is the easiest approach

to incorporate vegetation. It basically means

rasterising vegetation height information in case it is in

vector format and correcting the raster DEM with

vegetation heights by adding vegetation height values.

This approach does not take the nature of the

vegetation, its spatial distribution or density into

account. It is a relative easy to use approach and this

method has been used in several studies including

Domingo-Santos et al. (2011)s. The relative simple

approach to incorporate vegetation has been applied to

the SERF study area.

A few assumptions need to be taken into account

when adding a vegetation elevation model to the DEM

as mentioned by Domingo-Santos et al. (2011) and

Hernndez (2003). The main assumption is that

vegetation must be dense enough to be impenetrable

to sight because the possibility of seeing throughvegetation is not taken into account.

The vegetation maps for the Iron Age were based on a

generalised theory that many areas were deforested by

this time period, or at least the woodlands were

heavily managed. Therefore it is assumed that much

of the improved land that can be seen now would not

have large-scale forestry.

Only woodland has been modelled. Woodland would

have had consisted of mixed species, such as oak,birch, alder, hazel and pine. Heights of these trees vary

from 12 to 30 meters. Woodland would have been

partly managed in the Iron Age and are therefore

unlikely to have grown to the maximum height. An

average height of 15 meters has been set.

Figure 2. Iron Age vegetation layer. Yellow represents

Scrubs & Grassland, green represents Open Woodland, blue

represents Dense Woodland, rose represents Heather &Moor and purple represents Wetland & Bog.

A similar vegetation cover has been drawn for the

Neolithic. The modelling of vegetation was rather

difficult according to the archaeologists. There is little

evidence of tree cover in certain areas. This is not only

true for the SERF area but it is common in archaeology

that the patterns of vegetation can never be

reconstructed with any degree of precision, especially

for prehistoric times (Llobera, 2007).

The developed tool converts the polygon vegetation

cover into a raster, adds the height values of the raster


4/7

4

to the DEM raster values and creates a DEM with

vegetation incorporated. The corrected DEM can be

used as input for visibility analysis.

6. METHODOLOGY: adjusting theviewshed output

The main criticism on visibility analysis with GIS hasbeen with the binary output that does not reflect the

complexities of reality (Chapman, 2006). Fisher (1992)

tested a visibility tool in different GISs and came to the

conclusion that the results differ. Visibility analysis is

not precise, due to the sources of error listed before.

The Boolean representation (binary output) is

therefore not acceptable (Fisher, 1992). Alternative

methods have been proposed including generating

fuzzy viewsheds, where the degree of visibility is

being calculated (Fisher, 1992; Fisher 1993) and

probable viewsheds, where DEM uncertainty is takeninto account (Llobera, 2007; Fisher, 1992, 1993, 1994,

1998).

6.1Incorporating human constraintsVisibility is constrained by different factors such as

optical physics, atmospheric effects and psychological

and cultural factors (Ervin and Steinitz, 2003). Some of

these effects are integrated in the tools in GIS software

nowadays, such as atmospheric refraction (e.g. ArcGIS

9.3 has the option to check a box to use a refractivitycoefficient in some visibility tools).Distance is also afactor constraining visibility. Work has been done on

so called fuzzy viewsheds by (among others) Fisher(1995) and Ogburn (2006). These fuzzy viewsheds take

the effect of distance on visibility into account. The

underlying theory behind these fuzzy viewsheds is

that visibility decreases with distance from the

viewpoint. The resulting viewshed gives an indication

of the degree of visibility based on distance from the

viewpoint(s) with a lower degree of visibility with a

lower fuzzy membership. Fishers basic stepsinvolved, that have been adopted for this research, are

summed up below:

1. Creating a binary viewshed2. Creating a distance buffer around the

viewpoint(s)

3. Creating a distance decay buffer around theviewpoint(s) by applying a distance decay

function (see figure 3) to the distance buffer

4. Combining the distance decay buffer with thebinary viewshed

()

()

Figure 3. Fishers distance decay function, where:

= fuzzy membership, b1 = distance from viewpoint with

good visibility (foreground), b2 = distance from viewpoint at

which the visibility is 50%, = distance from theviewpoint to column 1 and row j

Figure 4. Distances b1 and b2. The outer circle represents

the cross-over point where visibility is assumed to be 50%

(source: Beaulieu (2007), figure 3)

6.2 Incorporating DEM uncertainty

Darnell at al. (2000) states that even a small amount of

elevation can greatly affect derivative products. When

using an ordinary viewshed tool, the accuracy of the

DEM will not be taken into account. Incorporating

DEM error will give a better representation of reality

to the end-user. The uncertainty of DEMs have been

discussed by multiple authors (among others Carlisle,

2002; Fisher, 1990, Darnell et al, 2007). Fisher (1992,1993, 1994, 1995, 2006) discusses the importance of

incorporating the uncertainty of DEMs for visibility

analyses.

The accuracy of the DEM is normally described with

the standard deviation or RMSE. The accuracy of the

DEM can be incorporated in viewshed analysis by

using a Monte Carlo approach where DEM values are

randomly changed based on their error. The Monte

Carlo simulation is based on the principle that the

DEM is one of an infinite number of possible

representations of reality (Carlisle, 2002). A number ofrealisations has to be chosen and for each of these

realisations a binary viewshed will be calculated. A

b2b1


5/7

5

probable viewshed will be the result with values

between 0 and the number of realisations. Values close

to the value of the number of realisations will have a

high probability of being visible. The basics steps

involved are summed up below.

1. Determine the required number of realisations2. Create random values for each cell in the DEM

with a normal distribution with a mean of 0and a standard deviation

3. Repeat creating random values for the DEMfor the number of realisations

4. Perform viewshed analysis for each of therealisations

5. Generate a probable viewshed by summingthe realisations

7. RESULTS AND DISCUSSIONFour enriched ArcGIS tools were created with the

ArcGIS ModelBuilder; Incorporate Vegetation,

Fuzzy Viewshed, Probable Viewshed and Combine

Fuzzy and Probable Viewsheds. These tools have been

used to analyse (inter)visibility of the ceremonial

complexes in the Neolithic and the forts in the Iron

Age.

7.1 Improving the DEM input

The main purpose of developing this tool was creating

a possibility to explore the effects of vegetation on

visibility. The approach used to accomplish this is a

relative simple approach but it does allow the user to

draw a vegetation model in the GIS and use this model

as an input vegetation layer to add to the DEM. The

number of visible cells from different features from the

Iron Age and Neolithic has been analysed.

Table 1. Effect of incorporating vegetation on the number of

visible cell calculated by using the ArcGIS 9.3 Spatial

Analyst Viewshed tool (single viewpoint). Iron Age fortsare the viewpoints.

Viewshed analyses with and without vegetation result

in a different number of visible cells (see table 1). The

number of visible cells is higher when viewshed

analysis is done without vegetation incorporated than

when viewshed analysis is done with vegetation

incorporated. This result is not surprising because

adding a vegetation layer increases the elevation

height of the cells with the height of the vegetation

cover (excluding the cells very near to the viewpoints,which are assumed to be cleared from vegetation).

This will cause less cells to be visible. From the data

shown in table 1, the incorporation of vegetation seems

to have the biggest impact on the viewshed from the

fort Law of Dumbuils, with 24% less cells visible. The

fort Law of Dumbuils is surrounded by open

woodland with a given screening height of 7.5m in the

vegetation model. This vegetation, relatively close to

the fort, blocks part of the view. Vegetation seems to

have a lesser impact on the viewsheds from the forts

Ben Effrey and Rossie Law. Ben Effrey and Rossie Laware also surrounded by woodland but these forts are

situated on high hills with a steeper slope. Vegetation

does not have a big influence on visibility here.

Similar analysis has been done for the Neolithic. The

discussed results are the outcomes of viewshed

analyses based on vegetation models. Other vegetation

models might be more accurate and/or appropriate.

As can be concluded from the above, vegetation does

have an influence on visibility. Archaeologists thinkthat this tool, despite being based on a simple

approach, is useful for exploring the effects that

vegetation can have on visibility. Improvements are

possible but it is a good starting point for the

incorporation of vegetation.

7.2 Adjusting the viewshed output

Fuzzy viewsheds - Incorporating human constraints

Fuzzy viewsheds incorporate a human perception

factor, human eyesight, that is causing uncertainty in

viewshed analysis. Making this uncertainty factor

visible to the user of viewshed analysis, gives an

insight to the uncertainty caused by this factor.

The visibility of cultural objects, however, not only

depends on the distance from the viewpoint and the

size of the object but also on the colour of the

background, the colour of the object and many other

factors. These factors are also applicable to the cultural

objects in the SERF area. For example, forts may havebeen made more or less visible by using distinctive

colours or colours that camouflage the fort.

Fort

viewpoint

(offset 2m)

Number of

visible

cells:

withoutvegetation

Number of

visible

cells: with

vegetation

Number

of visible

cells:

difference

Change

of

number

ofvisible

cells

Castle Law 792328 722403 69925 -9%

Dunknock 100466 81277 19189 -19%

Jackschairs

Wood

384647 322696 61951 -16%

Law of

Dumbuils

369421 280746 88675 -24%

Ben Effrey 278522 260064 18458 -7%

Rossie Law 1019303 963498 55805 -6%


6/7

6

Archaeologists think, speaking from experience, that

the formula and values used to calculate visibility

might be underestimating the limits of visibility for

some situations. It still gives an indication of the

degree of visibility depending on the distance from the

viewpoint and it makes archaeologists aware that

there is a limitation to human eyesight and that cells

further away are less likely to be seen by a human

viewer.

Probable viewsheds - Incorporating DEM uncertainty

Probable viewsheds incorporate the uncertainty of the

DEM into viewshed analysis. Visualising the effect of

the uncertainty of the DEM on visibility makes the

user aware of this effect. GIS users are often not aware

of or ignore the fact that a DEM has a certain degree of

uncertainty and that this uncertainty can have a

significant influence on their analyses.

Archaeologists find this tool useful because it makesthem think about DEM uncertainty and it gives them

an indication of where cells were highly likely to be

visible and where cells were less likely to be visible,

depending on the DEM error.

The standard deviation value used is an estimate of the

real standard deviation because the real standard

deviation is not reported by Ordnance Survey. The

used standard deviation value to generate a probable

viewshed will most likely be a good approximation of

the actual standard deviation because the standard

deviation of an area relatively close and comparable tothe SERF area has been used. Therefore the resulting

probable viewsheds will give a good insight in the

probability of cells being visible depending on DEM

uncertainty.

8. CONCLUSION AND FUTURE WORKViewsheds are the result of modelling. A model will

always be a representation of reality, rather than

reality itself. But therein also lies the strength of amodel. It is possible to generate possible outcomes

rather than realities. This is especially useful for

historical modelling because one does not exactly

know what the situation was like in the past.

Consideration of the theoretical issues associated

with GIS is a necessary precursor to wise use of the

technology in archaeological analysis. Wheatley and

Gillings (2000).

The incorporation of vegetation and the visualisationof uncertainty does lead to better visibility analysis for

landscape archaeology and the SERF project. The tools

give a better insight in the factors influencing visibility

and invite archaeologists to think through their

questions about the significance and meaning of

particular sites. The improved visibility tools could be

used at various stages of archaeological research; it

would not only be useful at the analysis stage but it

could also be very useful in the planning stage to

develop and refine research questions and to identify

areas that might have been of great importance.The relative simple approaches used for this research

give a good indication of the influence of the named

factors (vegetation, the distance between the observer

and the target and the DEM error) and the usage

results in a more critical approach to visibility analysis.

More research on factors influencing visibility and

how to integrate these in GIS could lead to an

improvement of the existing tools. More evidence

about past vegetation would help to refine the

vegetation model. Field observations could help assess

and improve the formula and values used to calculatethe distance that can be seen by an observer. The

determination of the exact DEM error would improve

the probable viewsheds. For the SERF project it is

recommended to carry out visibility analysis with

accustomed tools because they help to evaluate the

influence of factors on visibility which is not possible

with ordinary visibility analyses provided by GISs.

REFERENCES

ASHTON E., 2010, Viewshed Creation: From Digital Terrain

Model to Digital Surface Model, 2010 ESRI International User

Conference Paper UC 1193, July 12-16, San Diego

CARLISLE B., 2002, Digital Elevation Model Quality and

Uncertainty in DEM-based Spatial Modelling,

http://www.numyspace.co.uk/~unn_szbc1/PhD/ (accessed on

06-03-2012), PhD thesis, University of Greenwich

CHAPMAN H., 2006, Landscape Archaeology and GIS,

Gloucestershire: Tempus Publishing Ltd

FISHER P., 1990, Simulation of Error in Digital ElevationModels, Papers and Proceedings of the 13th Applied

Geography Conference, 37-43

DARNELL A. et al, 2007,A tool for assessing error in digital

elevation modelsfrom a users perspective, GISRUK 2007

Proceedings, Geographical Information Science Research

Conference, Ireland

DOMINGO-SANTOS J. et al, 2011, The visual exposure in forest

and rural landscapes: An algorithm and a GIS tool, Landscape

and Urban Planning 101, 52-28


7/7

7

ERVIN S. and C. STEINITZ, 2003. Landscape visibility

computation: necessary, but not sufficient, Environment and

Planning B: Planning and Design 30, 757-766

FISHER P., 1992, First Experiments in Viewshed Uncertainty:

Simulating Fuzzy Viewsheds, Photogrammetric Engineering

and Remote Sensing 58, 345-352

FISHER P., 1993,Algorithm and implementation uncertainty in

viewshed analysis, International Journal of Geographical

Information Science 7(4), 331-347

FISHER P., 1994, Probable and fuzzy models of the viewshed

operation, Innovations in GIS: selected papers from the First

National Conference on GIS Research UK, M.F.Worboys,

London, UK, Taylor and Francis, 161-175

FISHER P., 1998, Improved Modelling of Elevation Error with

Geostatistics, Geoinformatica 2(3), 215-233

FISHER P. and N. TATE, 2006, Causes and consequences of

error in digital elevation models, Progress in PhysicalGeography 30(4), 467- 489

GILLINGS M. and D. WHEATLEY, 2000, Vision, perception

and GIS: developing

enriched approaches to the study of archeological visibility,

Beyond the Map: Archaeology and Spatial Technologies.

Amsterdam, Netherlands, IOS Press: 1-27

GILLINGS M. and D. WHEATLEY, 2001, Seeing is not

believing: unresolved issues in archaeological visibility

analysis,Slapak, Bozidar (ed.) On the good use of

geographical information systems in archaeologicallandscape studies.Proceedings of the COST G2 Working

Group 2 round tableLuxembourg, Office for Official

Publications of the European Communities, 25-36.

HERNANDEZ J. et al, 2003, Estimating visual perception of

rural landscapes using GIS: the influence of vegetation. The case of

Esla Valley (Spain), Journal of Food Agriculture and

Environment 1(1), 139-141

KAY S. and T. SLY T, 2001,An Application of Cumulative

Viewshed Analysis, Archeologia e Calcolatori 12, 167-179

LLOBERA M., 2007,Modeling visibility through vegetation,

International Journal of Geographical Information Science

21(7), 799-810

OGBURN D., 2006,Assessing the Level of Visibility of Cultural

Objects in Past Landscapes, Journal of Archaeological Science

33, 405-413

Documents

CLGE SC2012_Alblas. the NETHERLANDS__Archaeological Visibility Analysis With GIS